- 27 Jun, 2019 3 commits
-
-
Hans Verkuil authored
In order to support multiple CEC devices for an HDMI connector, and to support cec_connector_info, drivers should use either a cec_notifier_conn_(un)register pair of functions (HDMI drivers) or a cec_notifier_cec_adap_(un)register pair (CEC adapter drivers). This replaces cec_notifier_get_conn/cec_notifier_put. For CEC adapters it is also no longer needed to call cec_notifier_register, cec_register_cec_notifier and cec_notifier_unregister. This is now all handled internally by the new functions. Signed-off-by:
Hans Verkuil <hverkuil-cisco@xs4all.nl> Signed-off-by:
Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
-
Dariusz Marcinkiewicz authored
Define struct cec_connector_info in media/cec.h and define CEC_CAP_CONNECTOR_INFO. In a later patch this will be moved to uapi/linux/cec.h. The CEC_CAP_CONNECTOR_INFO capability can be set by drivers, but cec_allocate_adapter() will remove it again until the public API for this can be enabled once all drm drivers wire this up correctly. Also add the cec_fill_conn_info_from_drm and cec_s_conn_info functions, which are needed by drm drivers to fill in the cec_connector info based on a drm_connector. The cec_notifier_(un)register and cec_register_cec_notifier prototypes were moved from cec-notifier.h to cec.h since cec.h no longer includes cec-notifier.h. These headers included each other before, which caused various problems. Due to these changes the seco-cec driver was changed as well: it should include cec-notifier.h, not cec.h. Signed-off-by:
Dariusz Marcinkiewicz <darekm@google.com> Signed-off-by:
-
Hans Verkuil authored
dev -> hdmi_dev conn -> conn_name Check if n->conn_name is not NULL before calling strcmp. Check the result of kstrdup, and clean up on error. Signed-off-by:
-
- 24 Jun, 2019 37 commits
-
-
Sakari Ailus authored
When Media Controller was merged to mainline long, long time ago, no-one bothered to think what its MAINTAINERS entry should be. Now that Media Controller is moved into its own directory, address this at the same time. So tell people to mail patches to myself and Laurent Pinchart. Note that the patches are still merged through the Media tree, just like any other driver or framework bits that have separate "mail patches to" entries different from the main drivers/media one. Signed-off-by:
Sakari Ailus <sakari.ailus@linux.intel.com> Signed-off-by:
Laurent Pinchart <laurent.pinchart@ideasonboard.com> Acked-by:
-
Dmitry Osipenko authored
Frequent IOMMU remappings take about 50% of CPU usage because there is quite a lot to remap. Defer dmabuf's unmapping by 5 seconds in order to mitigate the mapping overhead which goes away completely and driver works as fast as in a case of a disabled IOMMU. The case of a disabled IOMMU should also benefit a tad from the caching since CPU cache maintenance that happens on dmabuf's attaching takes some resources. Signed-off-by:
Dmitry Osipenko <digetx@gmail.com> Signed-off-by:
-
Dmitry Osipenko authored
All Tegra's could provide memory isolation for the video decoder hardware using IOMMU, it is also required for Tegra30+ in order to handle sparse dmabuf's which GPU exports in a default kernel configuration. Signed-off-by:
-
Hans Verkuil authored
This driver has three locking issues: - The wait_event_interruptible() condition calls hdpvr_get_next_buffer(dev) which uses a mutex, which is not allowed. Rewrite with list_empty_careful() that doesn't need locking. - In hdpvr_read() the call to hdpvr_stop_streaming() didn't lock io_mutex, but it should have since stop_streaming expects that. - In hdpvr_device_release() io_mutex was locked when calling flush_work(), but there it shouldn't take that mutex since the work done by flush_work() also wants to lock that mutex. There are also two other changes (suggested by Keith): - msecs_to_jiffies(4000); (a NOP) should have been msleep(4000). - Change v4l2_dbg to v4l2_info to always log if streaming had to be restarted. Reported-by:
Keith Pyle <kpyle@austin.rr.com> Suggested-by:
-
Boris Brezillon authored
cfg->type can be overridden by v4l2_ctrl_fill() and the new value is stored in the local type var. Fix the tests to use this local var. Fixes: 0996517c ("V4L/DVB: v4l2: Add new control handling framework") Cc: <stable@vger.kernel.org> Signed-off-by:
Boris Brezillon <boris.brezillon@collabora.com> [hverkuil-cisco@xs4all.nl: change to !qmenu and !qmenu_int (checkpatch)] Signed-off-by:
-
Eugen Hristev authored
Changed module parameters to static. Reported-by:
kbuild test robot <lkp@intel.com> Signed-off-by:
Eugen Hristev <eugen.hristev@microchip.com> Acked-by:
-
Ezequiel Garcia authored
Rework std_init adding an explicit initialization for compound controls. While here, make sure the control is initialized to zero, before providing default values for all its fields. Reviewed-by:
Ezequiel Garcia <ezequiel@collabora.com> Signed-off-by:
-
Ezequiel Garcia authored
Use the newly introduced vb2_get_buffer API and avoid accessing buffers in the queue directly. Signed-off-by:
-
Puranjay Mohan authored
Callers of hrtimer_forward_now() should save the return value in u64. change type of missed from unsigned long to u64. Signed-off-by:
Puranjay Mohan <puranjay12@gmail.com> Signed-off-by:
-
André Almeida authored
v4l2_fill_pixfmt() returns -EINVAL if the pixelformat used as parameter is invalid or if the user is trying to use a multiplanar format with the singleplanar API. Currently, the vimc_cap_try_fmt_vid_cap() returns such value, but vimc_cap_s_fmt_vid_cap() is ignoring it. Fix that and returns an error value if vimc_cap_try_fmt_vid_cap() has failed. Signed-off-by:
André Almeida <andrealmeid@collabora.com> Suggested-by:
Helen Koike <helen.koike@collabora.com> Signed-off-by:
-
Hans Verkuil authored
Instead of filling in the struct v4l2_capability device_caps field, fill in the struct video_device device_caps field. That way the V4L2 core knows what the capabilities of the video device are. But this only really works if all drivers use this, so convert all pci drivers in this patch. Tested with cx88-blackbird and ivtv PVR-350. Signed-off-by:
-
Janusz Krzysztofik authored
Extend parameter checks performed by v4l2_subdev_call() with a check for a non-NULL pad config pointer if V4L2_SUBDEV_FORMAT_TRY format type is requested so drivers don't need to care. Signed-off-by:
Janusz Krzysztofik <jmkrzyszt@gmail.com> Reviewed-by:
-
Janusz Krzysztofik authored
Parameters passed to check helpers are now obtained by dereferencing unverified pointer arguments. Check validity of those pointers first. Signed-off-by:
-
Janusz Krzysztofik authored
Correctness of format type (try or active) and pad number parameters passed to subdevice operation callbacks is now verified only for IOCTL calls. However, those callbacks are also used by drivers, e.g., V4L2 host interfaces. Since both subdev_do_ioctl() and drivers are using v4l2_subdev_call() macro while calling subdevice operations, move those parameter checks from subdev_do_ioctl() to v4l2_subdev_call() so we can avoid taking care of those checks inside drivers. Define a wrapper function for each operation callback in scope, then gather those wrappers in a static v4l2_subdev_ops structure so the v4l2_subdev_call() macro can find them easy if provided. Signed-off-by:
-
Philipp Zabel authored
Add support for dynamically changing the GOP size, bitrate, frame rate, constant intra quantization parameter, number of intra refresh macro blocks and slice mode parameters. Signed-off-by:
Philipp Zabel <p.zabel@pengutronix.de> Signed-off-by:
-
Philipp Zabel authored
Changing slice mode dynamically while encoding will require to calculate the register value again, so split it out into a separate function. Signed-off-by:
-
Philipp Zabel authored
Currently the platform device pointer is stored in struct coda_dev, only to convert it into a device pointer wherever it is used. Just store the device pointer directly. Signed-off-by:
Philipp Zabel <philipp.zabel@gmail.com> Signed-off-by:
-
Philipp Zabel authored
If reordering is not enabled, the last decoded frame has to be the last returned buffer. Otherwise wait for the firmware to report no more frame to display. In that case the return buffer is the last one as well, and can be reported as such. Signed-off-by:
-
Philipp Zabel authored
If there is still a buffer pending, mark it as the last buffer. It will create a meta that is flagged as last when the buffer is copied into the bitstream ring buffer. If there are no more buffers pending, find the last bitstream meta and mark it as last. If there is no bitstream meta either, wake up the capture queue as there will be no more decoded frames. Signed-off-by:
-
Philipp Zabel authored
Make sure that an encoder stop command running concurrently with an encoder finish_run always either flags the last returned buffer or wakes up the capture queue to signal the end of stream condition afterwards. Signed-off-by:
-
Philipp Zabel authored
Use the flagged last output buffer to also flag the corresponding capture buffer after encoding. This causes the end of stream event to be issued and the buffer to be dequeued with the last flag set. Signed-off-by:
-
Philipp Zabel authored
If there are no pending queued output buffers to be encoded, waking up the capture queue with -EPIPE signals end of stream. If there are pending buffers on the other hand, setting the V4L2_BUF_FLAG_LAST on the resulting encoded capture buffers is all that is needed. Signed-off-by:
-
Philipp Zabel authored
Using the output queue sequence counter to determine the last buffer to be encoded or decoded always was fragile at best. Now that we have the last buffer flag propagating from the output queue to the capture queue correctly, this is not needed anymore. Signed-off-by:
-
Philipp Zabel authored
Mark the last output buffer to be encoded. Signed-off-by:
-
Philipp Zabel authored
If there are still queued output buffers pending to be copied into the bitstream ring buffer, setting the stream end flag should be deferred until the marked last output buffer is written into the bitstream ring buffer. Signed-off-by:
-
Philipp Zabel authored
Mark the last output buffer to be decoded and only copy pending queued output buffers into the bitstream ring buffer in the BIT processor decoder case. Signed-off-by:
-
Philipp Zabel authored
Since V4L2_BUF_FLAG_LAST is a CAPTURE only flag, clear it from OUTPUT buffers in QBUF and DQBUF. This allows to use the flag internally to signal the last buffer to decode after a decoder stop command was issued. Signed-off-by:
-
Philipp Zabel authored
If no more frames are decoded in bitstream end mode, and a previously decoded frame has been returned, the firmware still increments the frame number. To avoid a sequence number mismatch after decoder restart, increment the sequence_offset correction parameter. Signed-off-by:
-
Philipp Zabel authored
The bitstream ringbuffer might be in an underrun state after draining, or it might still contain unread data if the previous decoder stop command was flagged as immediate. Flush the bitstream ring buffer during V4L2_DEC_CMD_START to get into a well defined state. Also fill the bitstream with buffers that have been queued during draining, to resume decoding immediately. Signed-off-by:
-
Philipp Zabel authored
On CODA960, sequence initialization can succeed if less than 512 bytes are ready in the bitstream ring buffer. On other variants, warn about too small payload in start_streaming. Signed-off-by:
-
Philipp Zabel authored
If the first buffer contains only headers, the sequence initialization command fails. On CodaHx4 the buffer must be padded to at least 512 bytes, on CODA960 it seems to be enough to just repeat the sequence and extension headers (MPEG-2) or the VOS and VO headers (MPEG-4) once for for sequence initialization to succeed without further bitstream data. On CodaHx4 the headers can be repeated multiple times until the 512 byte mark is reached. A similar issue was solved for h.264 by padding with a filler NAL in commit 0eef8940 ("[media] coda: pad first h.264 buffer to 512 bytes"). Signed-off-by:
-
Philipp Zabel authored
Pass vaddr and size to coda_bitstream_queue instead of a struct vb2_v4l2_buffer to make it reusable for queueing data that is not exactly a whole v4l2 buffer into the bitstream ringbuffer. Signed-off-by:
-
Philipp Zabel authored
Combine the separate auxiliary buffer, buffer meta, frame type, and decode error arrays into an array of struct coda_internal_frame. Signed-off-by:
-
Philipp Zabel authored
The stateful decoder API requires decoders to signal detection of stream dimensions via the V4L2_EVENT_SOURCE_CHANGE event. Signed-off-by:
-
Philipp Zabel authored
Add a sequence initialization work item to be run when OUTPUT buffers are queued in the initialization state. Signed-off-by:
-
Philipp Zabel authored
The sequence initialization already has to happen during the initialization phase, after headers have been queued on the OUTPUT queue. This means that sequence initialization has to be queued as a work item from QBUF on the OUTPUT queue. The internal framebuffer setup should be done later during VIDIOC_REQBUFS() on the CAPTURE queue. Signed-off-by:
-
Marco Felsch authored
When the DEC_CMD_STOP command is issued after the context has already consumed all the queued buffers, we need to make sure to wake the destination queue with last_buffer_dequeued set, to allow userspace to make progress in its EOS handling. As there might still be picture run workers pending at that point, we need to synchronize with them, so the sequence number comparison reads stable values. reword commit message] Signed-off-by:
Marco Felsch <m.felsch@pengutronix.de> Signed-off-by:
Lucas Stach <l.stach@pengutronix.de> Signed-off-by:
-
